Reported News
GPT-Rosalind-6: OpenAI's trusted-access biology model runs on GPT-6 Astra weights
GPT-Rosalind-6 is OpenAI's frontier reasoning model for biology, drug discovery and translational medicine, offered only through trusted access; its system card addendum, published October 30, 2026, says it "uses the same weights as GPT-6 Astra". OpenAI reports single-attempt success of 37.7% on GeneBench Pro v13, up from 16.7% for GPT-Rosalind-5.5.
- Changes reported
- 6
- listed below, each from a source we read
- We tested
- 0
- we did not run it ourselves
- Gaps we flag
- 4
- things we could not verify
- Sources
- 1
- 1 primary, linked below
What changed
| # | What the source says | Status |
|---|---|---|
| 01 | OpenAI's addendum says GPT-Rosalind-6 uses the same weights as GPT-6 Astra with no added training data or extra post-training; the change is fewer refusals on dual-use biology for vetted users. | Reported |
| 02 | On OpenAI's own life sciences benchmarks at the highest reasoning budget, it reports 49.6% on MedChemBench v7 (GPT-Rosalind-5.5: 28.9%) and 61.3% on LifeSciBench Gold (44.0%). | Reported |
| 03 | GeneBench Pro v13 single-attempt success is reported at 37.7% versus 16.7% for GPT-Rosalind-5.5, using a benchmark-specific configuration that OpenAI says differs from production. | Reported |
| 04 | OpenAI treats the launch as High capability in the biological and chemical domain, below Critical, and as Critical in cybersecurity, inheriting GPT-6 Astra's assessment there. | Reported |
| 05 | Access is limited to approved customers after review of mission, use case, governance and security; the addendum says this release does not broaden eligibility beyond earlier announcements. | Reported |
| 06 | Who it matters for: life sciences teams already approved for GPT-Rosalind Research. Everyone else cannot use it, and developers on GPT-6 Astra get no new model weights from this release. | Reported |
reported by the sourcemeasured or run by usnot verified
Primary source: OpenAI Deployment Safety Hub: GPT-Rosalind-6 System Card Addendum. We are the fast, independent index; the vendor’s own page is the source of record for this release.
In the source’s words
“GPT-Rosalind-6 uses the same weights as GPT-6 Astra. We did not add training data or perform additional post-training for this release.”
“At the highest reasoning budget tested, single-attempt success rates were 37.7% for GPT-Rosalind-6 and 16.7% for GPT-Rosalind-5.5.”
“GPT-Rosalind-6 is available only to approved customers.”
What we have not verified
- We did not run this release ourselves, and we cannot: it is available only to approved organizations.
- All benchmark scores are OpenAI's own, on benchmarks run by OpenAI; we found no independent results.
- The addendum gives a publication date (October 30, 2026) but no time of day, so the source publish time is left blank. We first saw it on November 2, about three days later.
- The addendum does not state pricing, an API model identifier or a context window; OpenAI's help center page on GPT-Rosalind blocked our automated fetch, so we have not used it as a source.
This brief is Reported: written from the sources below, not from running the release ourselves.
Sources
- OpenAI Deployment Safety Hub: GPT-Rosalind-6 System Card Addendumprimary source · fetched 2026-11-02
Related
Spotted an error? See our corrections policy or email the address on the contact page. Markdown version for agents: /news/gpt-rosalind-6-life-sciences-model.md
Want these briefs in your Google Top Stories? Add Agentic Weekly as a preferred source on Google.
The Weekly Verdict · free
5 new-model use cases. Every Monday.
A free email every Monday: the week's few big AI releases, then 5 short use cases for the new models, each with a prompt you can paste.
Free. Double opt-in. One click to unsubscribe.